BREAKING NEWS: Nvidia Crashes 16% on Chinese AI Breakthrough (DeepSeek)
25:52 Share

If you're a facilities manager at a university, you know students rely on the cafeteria for breakfast, lunch, dinner, and the occasional late night snack. So when a dishwasher breaks down and dirty plates pile up, the mess hall can turn messy in the blink of an eye. Enter Grainger. With over a million industrial grade products and fast delivery, the product you need now is never far away. So you can turn that dishwasher back into a lean, clean washing machine. Call

Click ranger.com or just stop by. Ranger, for the ones who get it done. At 1-800-Flowers.com, we know that connections are at the heart of being human. Whether celebrating life's joys or comforting during tough times, 1-800-Flowers helps you express what words can't. For nearly 50 years, millions have trusted 1-800-Flowers to deliver thoughtful gifts that help create lasting bonds.

Because it's more than just a gift. It's your way of showing you care. Visit 1-800-Flowers.com slash ACAST and connect today. That's 1-800-Flowers.com slash ACAST. Hey, I'm Ryan Reynolds. Recently, I asked Mint Mobile's legal team if big wireless companies are allowed to raise prices due to inflation. They said yes. And then when I asked if raising prices technically violates those onerous two-year contracts, they said, what the f*** are you talking about, you insane Hollywood a**hole?

So to recap, we're cutting the price of Mint Unlimited from $30 a month to just $15 a month. Give it a try at mintmobile.com slash switch. $45 upfront payment equivalent to $15 per month. New customers on first three-month plan only. Taxes and fees extra. Speeds lower above 40 gigabytes per details.

Welcome to the Market Maker podcast, hosted by me, Anthony Chung, where every Friday I talk to a member of the team about what happened in markets this week. From macro themes and single stock news to cryptocurrencies and careers in finance, our aim is simple, to make finance interesting and easy to understand for everyone. So let's get to it.

Hello and welcome to what is a breaking news episode. And the reason why is nothing like starting the week on a Monday morning with a big bang, a trillion dollar sell-off. You can already envisage the headlines at the end of today, Piers. But what's happened here is...

Something's come out over the weekend and it has seen the likes of the NASDAQ 100 in the futures market. The cash is going to open while we are live recording this and we'll update you. It's looking down about 5% with Nvidia down 12%.

And I was looking, quick look at the technicals on the actual NVIDIA chart. And it's going to open below quite a key phase of consolidation it's been in really over the last three months. So it could get pretty interesting throughout Monday's trading session for sure. But more than that, want to understand what is going on. What is this new Chinese AI startup breakthrough? Why is it so important? We're also going to look at why perhaps NVIDIA

a couple of things that you might want to consider if you were just kind of getting somewhat blindsided by what's going on. But Piers, to bring us up to speed, I know you had an important client call this afternoon, which kind of goes and testifies to how big this market and unforeseen the reaction has been because they've

They had a lot of senior traders on that call and they had to cancel it to go back to the desk, right? Well, I had a Morgan Stanley meeting with some sort of various kind of trading managing directors from various parts of the trading division. They've all postponed. They're like, we cannot leave the desk. It is absolutely kicked off.

to say the least. And actually, obviously there's a lot of trade volume going through. We'll talk about why in a sec. But as of this morning, and this was like, this is a few hours ago now, but 200,000 NASDAQ 100 futures contracts had traded, which was actually four times more than the 30-day average. And that was like one hour after the sort of open of that futures contract. So yeah, it's...

It's a classic scenario of people rolling out of bed and going, what, what, hang on, what? Are those numbers actually true? And I think it's been, it's almost like you can't say this wasn't sort of on the risk radar.

And it's always just one of those things that, okay, it's a risk, could happen, got no idea really when. Well, when is now? And so DeepSeek is the new name in town. And this is a Chinese mainland kind of startup. And I mean, this has been kind of rolling through, I'd say the second half of last week.

So they released a detailed paper. This is DeepSeek now. They released a detailed paper last week explaining how they basically had an instruction manual to build a large language model that could then automatically learn and improve itself. Remember that these large language models, let's say like Meta's Lama, for example, these large language models are key for training people

AI tools like ChatGPT, for example. And of course, up until this point, there's been a massive moat and a barrier to compete because of the cost of compute power.

NVIDIA, the poster child of the AI revolution, has an almost, you know, I'm not even sure what the percentage market share is, but it's almost the entire market share for producing these super chips that are these GPUs that are the only ones we thought that are powerful enough to kind of deal with all of this large language model training situation. And so, you know, and the US have been quite,

keen to protect their lead in this what is a technology arms race we'll come back to that point in a minute because some people are kind of

beginning to start to compare this to another arms race that happened a few decades ago. It kind of kicked off the space race, but we'll come on to that in a second. But yeah, basically, you know, NVIDIA, we had thought, had been well ahead of the game producing these chips. The Biden administration had prevented these chips from being exported to China. The idea being to hold China back so they can't compete on the same level. Well, this news...

really throws all of that out the window. And actually, maybe they can compete with insufferable chips, but also, as a result, on a budget that is a tiny fraction of what these big gun, you know, Mag-7 companies have been spending. And so, in short, this has thrown a spanner in the whole works of this AI revolution to date, where it's the rich...

And those big Mag7 companies that have got the cash to afford these NVIDIA chips, it was thought that it was an exclusive race. Not anymore. Just our episode last week, we were talking about how, you know, this kind of joint venture between OpenAI and SoftBank was like bringing together some entities to be able to provide the capital and firepower to get up onto the level to compete with,

you know, the Metas and the Googles and the Microsofts. And actually this news here is, well, actually forget all of that. You can do it at a fraction of the cost. So just looking, NVIDIA has just opened. It opened down, I think about 13% and it's just bounced back to down just 10 and a half now, but early days. Yeah. But just explain to me then, because Microsoft are down 4.4, Google's down 3.5. So all the big names, big tech. Yeah.

But just talk to me about the kind of context here, in terms of the implications on cost in particular. So the market reaction. All right. There's several different pockets of the market that are getting hammered here for slightly different reasons. So you've got the NVIDIAs. Well, firstly, so you've got your chip producers. You've got any companies that are supplying part of the AI supply chain.

So, NVIDIA, ARM, for example, they're your chip suppliers. ASML, sorry, they're getting hammered. And that's because, well, maybe you don't need their super advanced chips that are crazy expensive to get involved with this game, right? It was thought they had dominant market share. Maybe that market share is now going to get eroded. So that's kind of number one. Then you go back a step. Well, sorry, one more step through the supply chain. There's companies like Siemens Energy.

schneider electric they supply hardware for ai infrastructure right siemens energy german company down 22 today schneider's electric that's a french company they're down 9.2 right they're kind of electrical power products okay which are used in these big data centers okay so you've got those types of companies so they're directly involved with this ai software infrastructure side then

Take another angle. Well, all of these big mag seven companies, Meta, Microsoft, and actually Meta just on Friday, which was kind of after this kind of deep seek stuff started to move, Zuckerberg on Friday actually said, reconfirmed that Meta planned to spend $65 billion in 2025 on AI infrastructure. Microsoft last week said,

They said they're going to spend $80 billion in 2025 on infrastructure. You just had that OpenAI SoftBank tie up. They're going to spend $500 billion in the next five years, right? Now, up until this point, there's been a thesis that you have to spend big to be in this race.

Is there going to be a return on investment for, you know, are these companies that are spending big, are they the ones that are going to win the big pie that will be the revenue that comes out of all this AI world of the future, right? Well, all of this big news is basically potentially saying, well, you've wasted all that money. You shouldn't have spent it all. You can do it at a fraction of the cost. Okay, so that argument about

they're going to see big return on investment from these massive investments. That argument's been massively dented here. So you're getting the big companies, as you were mentioning there. So I'll just go through them. Yeah, Microsoft has just opened down 4.2%. You've got Meta down 2.5%. Google are down, all right. Not as much as NVIDIA, who are directly in the firing line, as in potential direct revenue hit here.

The second order effect is that lack of return on investment. And so Amazon are down 3% here, Tesla down 2% and so on, right? Then you've got, well, what does this mean for the big indices? Because obviously all these companies I've just mentioned, well, they're the big dominant forces in the indexes. So now you go and look at the NASDAQ, which has just opened, it's just opened down 3%, okay? And actually, it has bounced here a little bit.

So already, as you were mentioning, so that some of these, I mean, this is so live as we're talking, it was, I think we opened down about 5% on the NASDAQ, it's now down three. But of course, because these massive MAG7 companies make up such a huge part of the index, well, then, of course, they're getting hammered as well.

There's one Mag 7, your defensive Death Star, is actually up nearly 2%. So the fact that Apple is so far behind in the AI race has actually come as their savior today. You're absolutely right. Apple is up 1.6% here because they haven't spent anything yet or nowhere near the amount the others have. And it's almost like, well, hang on, that's genius.

Now they've held back. Well, now they can enter the race at a much cheaper price. Other market reactions, the VIX, which is looking at as kind of a volatility index, the fear gauge, we called it earlier on this morning, that was spiked up 45%. Let me see. I'll try and get a reading on the current.

price is now up 33%. So yeah, you're definitely getting some big kind of reversals. But these markets are all still quite heavily down, although the amount they're down is reducing as the more we go through this podcast, actually.

Yeah. And interestingly, this is earnings season. So we do, in fact, have the likes of Microsoft, Apple all coming out this week with their profit growth already under a bit of pressure. So interesting that they're obviously going to be asked.

in the analyst calls about this. And this has kind of shifted the entire focus on the, on the, you might remember a couple of years back, even before like three years back before the AI thing kicked off, um, meta we're talking about investing in the metaverse, right?

Right. And actually, as Zuckerberg came onto these earnings calls and said, we are going to spend X billions, X being a large number on this metaverse thing. It was almost like the more they were committing to spend, the worse it was. And their share price would get hammered. I think you might actually now get a beginning of that trend here.

for AI. So the more like if Microsoft on their earnings call later this week come out and said, yeah, we're still committed to $80 billion spend this year. I think you're going to see that feed through negatively in their share price. And analysts on the call are going to be asking, well, why? Now, why are you spending that amount of money? Surely this is a waste of cash flow and so on. So it'll be these earnings calls. I mean, the CFO, I do not

envy the CFOs and the CEOs on those earnings calls later this week having to try and navigate through this kind of very obviously live development. Yeah everyone's becoming an AI expert once again I'm sure the thought leadership on LinkedIn is going to be lit up later on today but it's going to be interesting what Trump says I mean look it's it's only 2 30 9 30 a.m over in the east coast of the US what's Trump's response obviously last week

Trump was like, you know, obviously slapping everyone on the back from OpenAI and SoftBank, coming to the Oval Office, aren't you amazing? We're going to spend 500 billion. We're going to make so, you know, create so many jobs. So the geopolitical angle here, you know, given that how the US had imposed those stringent restrictions on chip exports to China, thinking that they were hindering their ability to compete. I mean, well...

What do they say now? This is what happens when the president makes decisions without his co-president. If you leave Elon out in the cold, you have to pay a penalty, I'm afraid. But look, I just thought, you know, having worked in news aggregation, let's call it, in my career for 20 years,

There's definitely, I think, behavioral elements to this, i.e. the interpretation, as you would have seen many times, where it's that knee-jerk reaction, as you said at the top of the show. You wake up and you see the market's down that heavy. They start selling, momentum selling, and it all starts to trigger off. There's a couple of things I was just thinking, right, let's just step back and let's just ask a couple questions here about more legitimately whether this is a serious issue

kind of existential threat is kind of how the press are pinning it. And this is their job to talk it up. So a couple of things I'd just like to flag. One is the potential strategic messaging behind this. So i.e. a counter to the US leadership narrative of what we just had last week. So Deep Seek's announcement, I think,

It's not without surprise it comes just literal one week after this first week in office, this real kind of signposting of make America an AI leader and this investment project. And a couple of things on that. One,

It's Chinese New Year this week as well. So I think that timing is compounded, particularly because there was economic data that came out that again shows the Chinese economy is faltering. Industrial activity is weakening. There's still big question marks on many different pockets of the Chinese economy. And people in China are about to go and see their friends and loved ones all across the country and have a week of down tools.

And so that in combination with the Stargate announcement, you know, there's nothing like a little bolstering of national pride. And we see this with what we have with the quality of information.

So people always say you should always question the quality of Chinese economic statistics that you see. Why? Because there's such deep state kind of involvement in a lot of these things and deep seek self-censors on topics. So actually, if I was reading this thing, it was saying if you were to basically do a query on Tannerman Square protests,

or geopolitical events, it won't answer your questions. So actually then, yes, it's open source and all these different things, but it's being content curated in terms of what information is actually driving these models. So yeah, the DeepSeek bot is capable of giving detailed responses about political figures, of course, in India, but it won't say anything about Xi. So that's one thing. The next then is who's actually vetted this claim?

I mean, it seems quite clear. The case is very compelling and it's absolutely the perfect kind of antidote, if you like, for a Chinese response. But so far, there is no public third party confirmation of what DeepSeek are saying. Zero. And DeepSeek publication of the paper outlining their R1 model, while I think you can argue some transparency of putting that out does lend some credibility,

One of the things I'd be looking out for is independently verified experts who start running this model to see about its performance and efficiency claims. Because it's not like the Chinese don't overstate these things in the past. But to counter your point, they're at...

is right now available in the app store. And actually one of the things that's kind of triggered some of this reaction is it went to number one in the app store over the weekend. And so it's there, you can use their product. And actually people that are using it are saying, actually, wow, yeah, this is all right. It's not as quick and as good as the latest, you know, chat GPT version, but it's like, well, hang on, how have they built this when they have not had access to

to H100 chips from NVIDIA because they've been blocked because of sanctions. So it's like, well, how has it been possible for them to have done this

And sure, I mean, I don't know, I saw one number saying they managed to train it with a budget of $5 million. I don't know where that $5 million came from, rather than the multiple, multiple billions. I think actually, US big tech spent $250 billion last year on infrastructure, right? So $5 million, God knows where that number came from. But it's like, well, how have they done it? They've done it.

Now, how have they done it? Well, fine, they're spinning the story that they've done it on their own super cheap. Now, that hasn't been verified. I don't know how they've actually done it. But for now, the product is there. It's available. They can download it, have a go. And the experience is very good. Yeah, until Trump will come out and announce in the next 48 hours that that app is no longer going to be downloadable.

in certain geographic territories, I'm sure. But a couple of things then. So one, is this just a natural cycle? You said earlier, NVIDIA was so dominant in this space. Not that they're getting complacent, but you need an event like this in order to just sharpen the mind again from the US big tech scene

to just actually, this is a good wake-up call, talking not today, Monday's trading session, but over the more medium term, now that they're aware of this and the possibility of this, then that's a good thing. I mean, so there are counter-arguments. Well, firstly, hang on, before the counter-argument, it shouldn't be a surprise. I think it should be a surprise if the cost of all of this

remains crazy, ridiculously high. Prices set by NVIDIA because they've essentially cornered the market. There's this law. It's called Moore's Law. Basically, Microsoft wouldn't exist as a business if Moore's Law wasn't true.

So Moore's law is this law dates back to the 60s and the 70s and the 80s, where basically every two years, the cost of infrastructure for all things like using semiconductors and all the rest of it, anything computing, basically the cost of it all halves every two years.

This is Moore's law. And Microsoft in the 80s took advantage of computers going from being this massive, crazy, expensive things that take up a whole room to being on everybody's desk, you know, across the workforces of the Western world, right? That was only possible because the cost of it all dropped.

This is now, you know, why would you expect Moore's law to not apply in this new AI age? And I think some investors have been expecting it for some reason. They don't think it's going to apply. Well, here it is. And I don't think that should be surprising. That's one thing to say. But there are counter arguments, right? And look, you've got to be careful here because there's a lot of people talking their own book about

If you're still long NVIDIA, you're out, note to clients this morning going, ah, this is nonsense. You shouldn't believe it. You shouldn't believe these Chinese...

you know, unverified, you know, breaking news and what have you. One argument that's kind of coming to the surface of all of that, which might have some credibility, I would say, is that apparently they've, you know, DeepSeat were able to kind of take shortcuts in its own training costs by leveraging the latest models from OpenAI, right? Meaning, okay, they might have caught up. They're not

in line with OpenAI, and they're certainly not in front of them. They're behind, right? But they're a lot closer behind than we had thought. But if they're reliant on whatever latest model OpenAI has available out to the world, well, of course, they're always going to be behind.

Because OpenAI, well, they're working on their next version right now, right? It's not available to anybody other than themselves. So if they're reliant on OpenAI's models, they will always be 12 months behind. And so that's, I think, actually a pretty good argument to say that whilst this is, wow, shock, big news, it might not mean, it's not the death knell for NVIDIA's sort of, you know, 90% of their revenue.

And ultimately, if this was a completely Chinese-run shop and not using that open AI system, for example, surely the adoption of this AI would be impeded by a business's willingness to use that technology if it's Chinese-owned. Right.

Isn't it then better that this happens so big tech in the US can basically understand the kinks in their armor and understand the adoption techniques to drop costs, perhaps? Because there's no way that a Western developed world is going to buy Chinese and implement its technology within its business on a large scale, surely? No, but I think the point is that

If you're Microsoft sat there, right, you've budgeted out for 2025 and the biggest line item on your whole budget is 80 billion spend on AI infrastructure. If I was Microsoft, I'd now be going, okay, hang on a sec. Let's just revisit that line item. And how is it that this Chinese company...

has managed to do what they've done on a shoestring budget, actually, maybe we don't need to be spending disgusting amounts of billions. And maybe we can do this as in Microsoft, maybe we can do this much more efficiently. So I think it actually could be

In the medium term, could be good for the big tech as they actually end up spending a lot less than they would have done. In the short term, it's bad because they've already spent a ridiculous amount and it might be that it's wasted money. But then NVIDIA are right in the firing line because how much market share will they now lose if their chips are being sold at a crazy price and people don't need to buy them anymore? So I think...

This is going to be a messy few days and those earnings calls are going to be really key. NVIDIA right now is down 11%. Is NVIDIA going to rebound to unchanged and recover all of these losses? In my opinion, no.

I actually think for Nvidia, this is kind of end of chapter one where they've just had the whole market to themselves and they could price whatever they want. I think it's literally end of chapter one. Chapter two then, I'm not saying that now that's it, but it's just, it's going to be a step change and just an interesting turn of the page.

Yeah, and interestingly as well, to close, there was that consumer electronics show that we just had, what, three weeks ago? And NVIDIA were talking very much around robotics, autonomous vehicles. So are they looking to spread their wings as well into different parts of that AI cycle? Well, that's it. And that's what you should do. If you've got, in the good times, don't just sit back in your armchair on the beach and think, wow,

wow, aren't I amazing? You've got to start planning ahead, right? How are we going to diversify? You know, because they're obviously hugely reliant on one type of revenue. So it's about, right, let's diversify this thing out and make sure we're locked in, you know, across the different kind of stages of this sort of AI journey.

Okay, great. Well, look, great to get your thoughts. I know it's fresh off the press, so I'm sure there's gonna be lots of developments throughout the week. But yeah, good to get your take. And I hope everyone found that useful. Thanks, Piers. Thanks a lot.

If you're a facilities manager at a university, you know students rely on the cafeteria for breakfast, lunch, dinner, and the occasional late night snack. So when a dishwasher breaks down and dirty plates pile up, the mess hall can turn messy in the blink of an eye. Enter Grainger. With over a million industrial grade products and fast delivery, the product you need now is never far away. So you can turn that dishwasher back into a lean, clean washing machine. Call, click Grainger.com or just stop by.

Ranger, for the ones who get it done.

is here to solve that. Rocket Money is a personal finance app that helps find and cancel your unwanted subscriptions, monitors your spending, and helps lower your bills. You can see all those subscriptions that you've accrued over a lifetime of putting your credit card in on the internet in one place. If you don't want them, just cancel them. With a few taps, Rocket Money can help with that. Rocket Money has over 5 million users and has saved a total of $500 million in canceled subscriptions, saving members up to $740 a year when using all the app's premium features.

Stop wasting money on things you don't use. Cancel your unwanted subscriptions by going to rocketmoney.com slash cancel subs. That's rocketmoney.com slash cancel subs, not submarines.

Work takes up most of your time. That's why you should use stamps.com to save time with your mailing and shipping and have flexibility to focus on more important things. Stamps.com can handle all your mailing and shipping needs with rates up to 88% off USPS and UPS. Add flexibility to your day with stamps.com. Go to stamps.com slash program to sign up for a special offer. No

No contract. Cancel anytime. That's stance.com slash program.